A healthcare/philosophical question

[ianpnw]

How so? I'm willing to listen.


Edit: Fuck that. I'm not a customer at all. I'm a patient who knows what he wants.

The reason not to do random screening blood tests is a matter of math. If a test has 95% specificity, 5% of the normal population will have an “abnormal” result i.e. more than 2x SD from the norm. So if you do 2 random tests, only 90% of the population (0.95 x 0.95) will be within 2x SD from the norm, or in other words, 10% will have an “abnormal” result, even though they are not sick. If you do 10 random tests, 60% of the population (0.95 x 0.95 x 0.95 etc) will be within 2 x SD, and 40% will have an “abnormal” result, even though they aren’t sick. There are many studies from pre-op screening looking at just this issue, and a lot of unnecessary testing/treatment resulted from chasing spurious “abnormal” results.

[Toadman]

Next time bring a big wad of cash and just tell him to get the labs done. Or does that only work in the movies?

[Mazderati]

The reason not to do random screening blood tests is a matter of math. If a test has 95% specificity, 5% of the normal population will have an “abnormal” result i.e. more than 2x SD from the norm. So if you do 2 random tests, only 90% of the population (0.95 x 0.95) will be within 2x SD from the norm, or in other words, 10% will have an “abnormal” result, even though they are not sick. If you do 10 random tests, 60% of the population (0.95 x 0.95 x 0.95 etc) will be within 2 x SD, and 40% will have an “abnormal” result, even though they aren’t sick. There are many studies from pre-op screening looking at just this issue, and a lot of unnecessary testing/treatment resulted from chasing spurious “abnormal” results.

I am a patient, not a population. Retest to confirm.

[jono]

Go see someone else.

^^QFT because you know:

Demand is a loaded word in this context. Patients should be able to request common labs and a doctor should be willing to order them as part of collaborative, preventive care. Missing a true concern because a lab was not ordered is also a risk. Discuss caveats before and interpretations after resulting. I appreciate the perspective.

Guidelines are flexible. They are not rules. But, they do often consider cost.

An erroneous result does not automatically mean out of range or indicate invasive diagnostics. It might just mean retesting or keeping an eye on it at the next annual. Labs are information. What to do with that information is where medical training and experience are invaluable.

Is the public health burden of early or late disease detection higher?

"I don't want my patient to know because it might make them feel some kind of way," is next-level preventive medicine.

...and if your doc thinks you don't know all that ^^ then your greatest risk is miscommunication going forward. Which can be deadly when the topic is ruling out potentially life-threatening diseases. If he thinks you're a hypochondriac who's always angling for tests he's going to resist, argue or prejudice himself.

Being on the same page with your doc is very important. Certainly don't accept one who interferes with your ability to be your own advocate when that's needed.

Anecdotally (bias recognition): I left my last doc after an egregious ~8 month miscommunication/delayed diagnosis. I've been seeing a PA (former employee of his) ever since because she knows the story so we never assume we know the other's thoughts. Femaleness might help on that front, but we both expect detailed discussion, research and follow up. Maybe PA-ego helps with collaboration like it does with referrals, too.

[Trackhead]

Per old goat: ordering tests based on suspicion

I had a post chemo pet scan with some abnormal uptake in a lymph node not previously affected by lymphoma. My oncologist didn’t think it was recurrence but offered me a biopsy if I was scared. I asked her for her professional opinion, which was, “I don’t think this represents lymphoma”.

I didn’t get the biopsy, follow up scans were all negative.

Sometimes tests are too sensitive and lead to invasive procedures, anxiety, etc. Think PSA and prostate biopsy….

Anyway, sorry for the anecdotes. If you want routine labs I’d order them for you and just let you know about current recommendations and call it a day. “Shared decision making” is the verbiage in the chart. Covers my ass.

[old goat]

Is this a joke on confirmation bias? Over my head if not.

No not a joke. Every test has a risk of false positives and false negatives. If you do the test on people in whom you expect the disease, the chances that a positive is a false positive is much lower than if you do the test on presumably healthy people. Likewise if you do the test on healthy people the chance that a negative is a false negative is much lower than if you do the test on sick people. The rate of false positives and false negatives depends on what population the test was done on.

The reason not to do random screening blood tests is a matter of math. If a test has 95% specificity, 5% of the normal population will have an “abnormal” result i.e. more than 2x SD from the norm. So if you do 2 random tests, only 90% of the population (0.95 x 0.95) will be within 2x SD from the norm, or in other words, 10% will have an “abnormal” result, even though they are not sick. If you do 10 random tests, 60% of the population (0.95 x 0.95 x 0.95 etc) will be within 2 x SD, and 40% will have an “abnormal” result, even though they aren’t sick. There are many studies from pre-op screening looking at just this issue, and a lot of unnecessary testing/treatment resulted from chasing spurious “abnormal” results.

What he said. A corollary is--the accepted threshold for deciding a treatment is beneficial is that statistical analysis shows that there is a 5% chance that the study is wrong. In statistical notation p<0.05. Which means that potentially 5% of studies are wrong, which is a small percent of a very large number--potentially thousands of wrong studies. In practice most studies that show a benefit are significant at a much lower p value than 0.05 so the number of false studies is lower, but still large.

I am a patient, not a population. Retest to confirm.

Retesting only helps if the problem is lab error, not if the problem is anomalous physiology.
Like it or not, testing on populations is the only way to develop evidence based medicine. We can try to narrow down populations to determine whether a treatment is more beneficial for certain subpopulations--women for example, of people over a certain age. But we are unable to narrow it down to the individual level, not yet at least. That may not matter if we're talking about cheap lab work with zero direct harm, other than the risk of chasing a positive result. It matters more and more as the test gets expensive, the medication has side effects, or the operation can kill you. Oh, and when I was a med student one of the hospital ward clerks had a CBC done for a cold, got a severely infected arm, lost some tissue, and was in the ICU septic for a while. If we're tossing around anecdotes.

I hope no one reading this is interpreting it mean that no one should have tests or that Mazderati's doc is right not to order any tests. Obviously healthy people should have some screening tests. The questions is which ones and how often and that is determined by risk factors like age, smoking, gender, obesity, family history, alcohol use, etc etc.

Here's a slightly related provider perspective on routine testing. The average breast cancer is considered to have been present for 9 years before it is detected. A smart lawyer started advertising in the newspaper (this was a long time ago) for women with breast cancer. His theory was, if the cancer was present for 9 years, any mammogram the woman had before the cancer was detected "missed" the cancer and therefore that was malpractice. I don't know how far he got with this approach.

[jono]

Worst case, a physiological anomaly should (mostly) be an argument against a first test, though, not a reason to stop doing tests where there's already a history.

The downsides of statistics aren't really an argument against treating a patient as an individual, either. I don't want to be treated in the blind, for example. One of the numerous factors that large studies seek to eliminate in order to target a specific factor might be the exact thing I need to find out as an individual.

That's an interesting point about breast cancer. Do you happen to recall what the 9 year estimate was based on? Just average mass and cellular growth rates, or are there more specific ways to look at that?

[old goat]

Worst case, a physiological anomaly should (mostly) be an argument against a first test, though, not a reason to stop doing tests where there's already a history.

The downsides of statistics aren't really an argument against treating a patient as an individual, either. I don't want to be treated in the blind, for example. One of the numerous factors that large studies seek to eliminate in order to target a specific factor might be the exact thing I need to find out as an individual.

That's an interesting point about breast cancer. Do you happen to recall what the 9 year estimate was based on? Just average mass and cellular growth rates, or are there more specific ways to look at that?

I can't agree with you about treating you as an individual, unless there is some evidence to support treating someone with your particular profile. If we are going to say the data tells us that data shows the treatment doesn't help but we're going to treat you because it might work on you--we might as well stop doing randomized double blind controlled trials and just guess at what might work. I'm not talking about a terminal disease where there is a 5% chance the treatment will help--if you want that chance, go for it. I'm talking about situations like for example, a nonfatal chronic disease where the studies show the treated group does not better or even worse than the placebo group. If you can isolate a subgroup that does better and you're in it, fine, go ahead and treat.

Which brings up another thing about trials--pretrial vs posttrial randomization. If your hypothesis is that the treatment might work better on women than on men, for example, and you plan to compare the results in the two groups and find a benefit in men, fine. But if you do a trial and the treatment doesn't work, you can't go back and try to find a subgroup that did better. By chance alone, you will usually be able to find one. That result is only valid if you do another trial on that subgroup and confirm that it really works on them.

We all want to be acknowledged as individuals, especially by our doctors, but that mainly applies to our social circumstances, our life preferences, and the likely. If we're white, male, 74 year old hypertensives with a history of coronary artery disease we're pretty much the same as all the other people who meet that description. If I want a stent so I can keep skiing, vs medication that might make it harder to ski, fine, give me the stent. But that preference doesn't change the chance of the stent working or the risk of complications from it.

[jono]

I'm not advocating for ignoring evidence, I'm saying that there's a difference between what we do when we're gathering evidence and how we use it. More precisely, the entire purpose of statistical analysis of a population is to assess the likelihood that a given treatment or factor will be effective for an individual. We hope the group is representative of the individual but the study can't know that. So it is up to the doctor seeing the individual to do the work to know if that person stands apart in meaningful ways.

Since not every meaningful factor gets studied, and many are intentionally ignored in order to get statistical significance without an even bigger sample, a lack of evidence alone is not always conclusive. (As a surgically-modified individual I would much prefer to be in the middle of the curve: when the population is too small to study others we don't get to throw up our hands and do nothing for lack of evidence. )

Best case, stats don't give an answer, only a probability. Statistics are the lowest form of math, but sometimes they're all we have, so we just have to remember their limitations. My point above was that the "subjective"-feeling aspect of this is a common barrier to communication. We've demonstrated that a bit, I think.

Lots of people understand and use statistics in different ways. I stopped meeting pre-med students in math classes after freshman year, so I imagine our uses diverge a bit. As that happens a lot of people tend to assume the other person is wrong (maybe justifiably, since wrong is common). But it's hard to know what someone else knows. It often takes time and nuanced discussion to make sure you're on the same page. If you can confirm that your aren't and won't be, at least there's an obvious solution.

[old goat]

The argument in your first paragraph has been used and is still used by doctors to justify lucrative procedures that have been shown to have no benefit--well, it might work for this individual, it's my gut feeling. Which is how we wound up with tedious and error prone pre authorization requirements from insurance companies.

Your valid point is that the doctor has to determine if the individual differs from the population in meaningful ways. But the catch here is --what is meaningful. I say it is a difference that has been shown in studies to change the outcome of a treatment, or at the very least, a difference that can be scientifically explained as likely to improve the outcome of a procedure. If a study done only on men shows a treatment doesn't work and your patient is a woman, can you explain why being a woman would improve the chance the treatment would help? Being different is not enough. (Big issue--too many studies in the past were only done in men. Women are nearly a separate species--after all women and men have different genomes. They deserve their own studies or at least the two main genders --that's an amusing phrase I read the other day--should be included in every study.)

It's tough to talk about this stuff in the abstract. It would be better if we could talk about a particular treatment for a particular problem that someone feels they were unjustly denied because of lack of proven benefit.

[jono]

My first paragraph was not advocating for gut feelings.

But yeah, I got the feeling we weren't discussing the same thing. The specifics in Mazdarati's post are mainly what I focused on, not surgery. Detection tools that cost little and are pretty noninvasive.

The understanding of which certainly benefit from some advanced understanding of stats because, in many cases, there are weird findings out there. As an example, I recall learning of a correlation between a first-time high blood sugar result and cancer--like 10% IMS. Not a huge flashing red light on its own, but should be enough to prick up the ears of someone dealing with one of the many sets of symptoms that can have cancer as an explanation. Enough to justify some tests.

Similar to cheap tests, I discuss a lot of treatments with my PA which are relatively low cost and worth a try from a risk management perspective. (Where risk = cost x probability, not any nebulous rounding down of every unlikelihood.) A ton of things discussed at an annual are not expensive procedures--or, better, they're alternatives to try to avoid those.

[XXX-er]

How many hypochondriacs do y'all got down there ?

My last wife was getting a prescription filled on average every 4 days according to her 3rd husband who figured this out cuz he got the drug bills, he just couldn't take a Hypodhindriac anymore and just didnt come home one day

[old goat]

My first paragraph was not advocating for gut feelings.

But yeah, I got the feeling we weren't discussing the same thing. The specifics in Mazdarati's post are mainly what I focused on, not surgery. Detection tools that cost little and are pretty noninvasive.

The understanding of which certainly benefit from some advanced understanding of stats because, in many cases, there are weird findings out there. As an example, I recall learning of a correlation between a first-time high blood sugar result and cancer--like 10% IMS. Not a huge flashing red light on its own, but should be enough to prick up the ears of someone dealing with one of the many sets of symptoms that can have cancer as an explanation. Enough to justify some tests.

Similar to cheap tests, I discuss a lot of treatments with my PA which are relatively low cost and worth a try from a risk management perspective. (Where risk = cost x probability, not any nebulous rounding down of every unlikelihood.) A ton of things discussed at an annual are not expensive procedures--or, better, they're alternatives to try to avoid those.

Yeah, I pivoted to drugs and operations because the consequences are much more significant and the discussion more interesting than whether someone gets a cheap blood test.

Weird statistics: some years ago there was a study that showed if a woman got a breast biopsy midcycle it was more likely to be cancer than if she got the biopsy when she was menstruating. This caused quite a stir until a second study showed the opposite. Combine the two and it made no difference when the biopsy was done. That's why we see meta=analysis, combining the results of multiple studies.

And don't get me started on relative vs absolute benefit.

[jono]

Numeracy has a very high standard deviation, with all the misunderstandings that accords, but relative vs. absolute might be the strongest example.